Abstract: | The well-known Q coefficient of association defined by YULE (1900) for two attributes is generalized from first-order association Q to second-order association Q for three attributes and to higher-order Q's for more than three attributes (or binary variables). Q analysis of t binary variables is shown to be a suitable means for describing simple (quasi-parametric) or complex (multi-nonparametric) relationships in a t-dimensional contingency cube. A biometrical example is given using clinical data. |